|
|
void | magma_get_zpotrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| | Returns in nb and recnb the crossover points for potrf based on n.
|
| |
| void | magma_get_cpotrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| |
| void | magma_get_dpotrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| |
| void | magma_get_spotrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| |
|
void | magma_get_zgetrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| | Returns in nb and recnb the crossover points for getrf.
|
| |
| void | magma_get_cgetrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| |
| void | magma_get_dgetrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| |
| void | magma_get_sgetrf_batched_nbparam (magma_int_t n, magma_int_t *nb, magma_int_t *recnb) |
| |
|
void | magma_get_zgetrf_vbatched_nbparam (magma_int_t max_m, magma_int_t max_n, magma_int_t *nb, magma_int_t *recnb) |
| | Returns in nb and recnb the crossover points for getrf.
|
| |
| void | magma_get_cgetrf_vbatched_nbparam (magma_int_t max_m, magma_int_t max_n, magma_int_t *nb, magma_int_t *recnb) |
| |
| void | magma_get_dgetrf_vbatched_nbparam (magma_int_t max_m, magma_int_t max_n, magma_int_t *nb, magma_int_t *recnb) |
| |
| void | magma_get_sgetrf_vbatched_nbparam (magma_int_t max_m, magma_int_t max_n, magma_int_t *nb, magma_int_t *recnb) |
| |
| magma_int_t | magma_get_zgeqrf_batched_nb (magma_int_t m) |
| |
| magma_int_t | magma_get_cgeqrf_batched_nb (magma_int_t m) |
| |
| magma_int_t | magma_get_dgeqrf_batched_nb (magma_int_t m) |
| |
| magma_int_t | magma_get_sgeqrf_batched_nb (magma_int_t m) |
| |
| magma_int_t | magma_get_zgeqr2_fused_sm_batched_nthreads (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zpotrf_batched_crossover () |
| |
| magma_int_t | magma_get_cpotrf_batched_crossover () |
| |
| magma_int_t | magma_get_dpotrf_batched_crossover () |
| |
| magma_int_t | magma_get_spotrf_batched_crossover () |
| |
| magma_int_t | magma_get_zpotrf_vbatched_crossover () |
| |
| magma_int_t | magma_get_cpotrf_vbatched_crossover () |
| |
| magma_int_t | magma_get_dpotrf_vbatched_crossover () |
| |
| magma_int_t | magma_get_spotrf_vbatched_crossover () |
| |
| magma_int_t | magma_get_zgetri_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgetri_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgetri_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgetri_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_ztrsm_batched_stop_nb (magma_side_t side, magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_ctrsm_batched_stop_nb (magma_side_t side, magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dtrsm_batched_stop_nb (magma_side_t side, magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_strsm_batched_stop_nb (magma_side_t side, magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgemm_batched_smallsq_limit (magma_int_t n) |
| |
| magma_int_t | magma_get_dgemm_batched_smallsq_limit (magma_int_t n) |
| |
| magma_int_t | magma_get_sgemm_batched_smallsq_limit (magma_int_t n) |
| |
| magma_int_t | magma_srecommend_cublas_gemm_batched (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| | Decides which is better (magma or cublas_batched), regardless of the performance of cublas stream. More...
|
| |
| magma_int_t | magma_drecommend_cublas_gemm_batched (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| |
| magma_int_t | magma_crecommend_cublas_gemm_batched (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| |
| magma_int_t | magma_zrecommend_cublas_gemm_batched (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| |
| magma_int_t | magma_srecommend_cublas_gemm_stream (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| | Decides if cublas stream should be used for a given gemm dimension/shape. More...
|
| |
| magma_int_t | magma_drecommend_cublas_gemm_stream (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| |
| magma_int_t | magma_crecommend_cublas_gemm_stream (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| |
| magma_int_t | magma_zrecommend_cublas_gemm_stream (magma_trans_t transa, magma_trans_t transb, magma_int_t m, magma_int_t n, magma_int_t k) |
| |
| magma_int_t | magma_get_spotrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dpotrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_cpotrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zpotrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zpotrf_right_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_cpotrf_right_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dpotrf_right_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_spotrf_right_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_sgeqp3_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgeqp3_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgeqp3_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgeqp3_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgeqrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgeqrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgeqrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgeqrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgeqlf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgeqlf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgeqlf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgeqlf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgelqf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgelqf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgelqf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgelqf_nb (magma_int_t m, magma_int_t n) |
| |
| double | magma_get_gemex_rankk_time (magma_int_t m, magma_int_t k, magma_mp_type_t gmtype) |
| |
| magma_int_t | magma_get_sgetrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgetrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgetrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgetrf_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgetrf_native_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgetrf_native_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgetrf_native_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgetrf_native_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgehrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dgehrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_cgehrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zgehrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_ssytrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dsytrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_chetrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zhetrd_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zhetrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_chetrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dsytrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_ssytrf_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zhetrf_aasen_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_chetrf_aasen_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dsytrf_aasen_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_ssytrf_aasen_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zhetrf_nopiv_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_chetrf_nopiv_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dsytrf_nopiv_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_ssytrf_nopiv_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_sgebrd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgebrd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgebrd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgebrd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_ssygst_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dsygst_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_chegst_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zhegst_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_sgetri_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dgetri_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_cgetri_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zgetri_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_sgesvd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgesvd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgesvd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgesvd_nb (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_ssygst_m_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dsygst_m_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_chegst_m_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zhegst_m_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_sbulge_gcperf () |
| |
| magma_int_t | magma_get_dbulge_gcperf () |
| |
| magma_int_t | magma_get_cbulge_gcperf () |
| |
| magma_int_t | magma_get_zbulge_gcperf () |
| |
| magma_int_t | magma_get_smlsize_divideconquer () |
| |
| magma_int_t | magma_get_sbulge_nb (magma_int_t n, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_dbulge_nb (magma_int_t n, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_cbulge_nb (magma_int_t n, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_zbulge_nb (magma_int_t n, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_sbulge_vblksiz (magma_int_t n, magma_int_t nb, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_dbulge_vblksiz (magma_int_t n, magma_int_t nb, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_cbulge_vblksiz (magma_int_t n, magma_int_t nb, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_zbulge_vblksiz (magma_int_t n, magma_int_t nb, magma_int_t nbthreads) |
| |
| magma_int_t | magma_get_sbulge_mgpu_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_dbulge_mgpu_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_cbulge_mgpu_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zbulge_mgpu_nb (magma_int_t n) |
| |
| magma_int_t | magma_get_zgemm_batched_ntcol (magma_int_t m) |
| |
| magma_int_t | magma_get_cgemm_batched_ntcol (magma_int_t m) |
| |
| magma_int_t | magma_get_dgemm_batched_ntcol (magma_int_t m) |
| |
| magma_int_t | magma_get_sgemm_batched_ntcol (magma_int_t m) |
| |
| magma_int_t | magma_get_zgetrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgetrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgetrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgetrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_zgeqrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_cgeqrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_dgeqrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
| magma_int_t | magma_get_sgeqrf_batched_ntcol (magma_int_t m, magma_int_t n) |
| |
Optimal block sizes vary with GPU and, to a lesser extent, CPU.
Tuning of the batched kernels that are invoked on extremely small sizes.
Kepler tuning was on K20c 705 MHz with SandyBridge 2.6 GHz host (bunsen). Fermi tuning was on S2050 1147 MHz with AMD Opteron 2.4 GHz host (romulus).